Accelerated Monte-Carlo Techniques for Modeling of Chain Architecture and Semi-Batch Radical Polymerization Process Optimization
This thesis introduces a new strategy to develop a comprehensive stochastic polymerization simulation with the capacity of considering the secondary reactions attributed to the radical acrylate/methacrylate/styrene polymerization carried out at high temperatures via semi-batch reactor operation. Unlike the conventional deterministic method of polymerization modeling (i.e., method of moments), the resulting kinetic Monte Carlo (KMC) model can predict the distribution of comonomer units among polymer chains. A combination of acceleration methods is implemented to effectively reduce the computational cost of these simulations while preserving the accuracy of the solution. KMC model output is not only compared to a deterministic model implemented in Predici® but experimentally challenged by comparing predictions to the properties of polymer samples extracted after crosslinking of resin synthesized in a set of 2-hydroxyethyl acrylate (HEA)/butyl methacrylate (BMA) copolymerizations. While the simulation time has been greatly reduced, the KMC method is still computationally costly relative to deterministic methods. Thus, a methodology has been developed to use the instantaneous copolymer composition and number-average chain lengths output from a deterministic model to estimate the instantaneous mole and weight fractions of polymer chains with respect to the number of comonomer units they possess, with the cumulative quantities calculated through suitable integration. Derivative-free optimization algorithms (e.g., Pattern Search and Particle Swarm) are combined with the accelerated kinetic Monte Carlo model to develop strategies to improve the traditional starved-feed policy often used by industry. Starved-intervals are introduced as an approach to modify the monomer and initiator feeding strategy. This concept is demonstrated using a simpler deterministic model based on the method of moments as a case study. A feeding schedule formulated by utilizing the Pattern Search optimization algorithm combined with the KMC model demonstrates remarkable improvement compared to the traditional starved-feed policy; in fact, this nonlinear feeding strategy reduces the total reaction time from 6 hours to less than 2 hours while the quality of polymer product is improved. This strategy is experimentally verified both at Queen’s University and Axalta Coasting Systems research facility.
URI for this recordhttp://hdl.handle.net/1974/27830
Request an alternative formatIf you require this document in an alternate, accessible format, please contact the Queen's Adaptive Technology Centre
The following license files are associated with this item: